CAMINO - Contextually Aware Mediation of Intent for Network Orchestration
A network configuration change has to make sense for the operator’s goals and the conditions around the network. CAMINO brings that context into the assessment of each proposed action.
The same action can have a different meaning in context
Autonomous network applications can pursue sensible goals while proposing incompatible changes. A prediction that one performance indicator will improve does not settle whether the action serves the operator’s current priorities.
In CAMINO, we assess individual configuration proposals using predicted network effects, operator intent and information about the surrounding environment. The architecture can reject a particular action while leaving the proposing application available to make other useful changes.
See why a favourable prediction can lead to rejection
Read the proposal in context
Predicted KPI effects
Favourable overallAssess predicted changes against the operator’s targets and KPI priorities.
Traffic context
Small favourable contributionLocal traffic conditions contribute to the assessment of this action.
Weather context
Substantial adverse contributionAnomalous weather adds a negative effect on accessibility KPIs.
In the published example, the adverse weather contribution leads to rejection despite favourable overall KPI predictions and a small positive traffic contribution.
Based on the explanation accompanying Figure 7 of the published paper. The cards show the direction and qualitative significance of reported contributions; they are not numerical scores or a live implementation.
From a proposal to a contextual decision
- 01Compare similar cells
Group network entities by configuration and operating conditions.
- 02Predict KPI effects
Estimate the effects associated with the proposed parameter change.
- 03Apply intent and context
Consider targets, priorities, weather, traffic and relevant operating information.
- 04Assess the action
Use the contextual assessment to approve or reject the individual proposal.
The KPI Degradation Prediction Layer supplies predictions. The Intent-Aware Conflict Detection component considers them alongside operator targets and contextual inputs. An intent interface supports targets that vary by area and time. The demonstrated decision rules use domain knowledge about how particular changes interact with the environment.
We illustrated the architecture using proposed configuration changes for an 800-cell LTE network and weather and traffic data collected through external APIs. The examples include power settings, target BLER and cell radius. They show how the different contributions produce a decision that can be explained to the operator.
Architecture and scenario evidence
The paper presents an architectural proposal and scenario-based assessments. It does not report verified gains from a production deployment or a head-to-head prediction-accuracy benchmark against other conflict-management schemes.
The demonstrated rules are manually defined. Their quality depends on the predictions and contextual data supplied to them. External data can be delayed or missing, clustering can change, and a deployment must account for the time needed to ingest data and assess a proposal. Adaptive rule learning and production validation are further work.
Conflict detection, arbitration and enforcement
CAMINO extends the trajectory of pre-emptive conflict detection and cell similarity. TACIT later examines how prediction accuracy can determine an application’s influence during arbitration.
ORACLE addresses verification of the arbitration workflow through a shared ledger. These pages describe related approaches with different mechanisms and evaluations. They do not establish the performance of a combined CAMINO–TACIT–ORACLE system.
Paper and citation
J. Armstrong, E. Fallon and S. Fallon, “CAMINO - Contextually Aware Mediation of Intent for Network Orchestration,” Telecommunication Systems, vol. 88, article 96, 2025. DOI: 10.1007/s11235-025-01324-9.